ERP for semiconductor fabs cannot be separated from the equipment automation layer. Unlike generic manufacturing ERP that manages materials by aggregated lot, fab ERP must track the exact position and history of every single wafer as it moves through hundreds of processing steps — and must know when a preventive-maintenance schedule should be deferred because a station is congested, or pulled forward to exploit a gap in the flow.
This is a problem fab ERP solves differently from ERP in any other industry: not just asking "is this lot ready," but asking "if we pull this tool for maintenance right now, can the remaining tools at this station still meet the 14-day production target."
SECS/GEM and GEM300 — the communication foundation fab ERP must integrate
SEMI E30 (GEM) is the standard implementation of SECS-II (SEMI E5) for all semiconductor manufacturing equipment, defining a common set of equipment behavior and communication capabilities, referencing SEMI E4 (SECS-I) and SEMI E37 (HSMS). Fab ERP doesn't read raw SECS-II directly but through the higher-level GEM300 family: SEMI E87 (Carrier Management) automatically authenticates FOUP carriers, SEMI E90 (Substrate Tracking) tracks the position and history of every individual wafer through each processing chamber, SEMI E94 (Control Job Management) converts enterprise production plans into automated commands sent down to station controllers.
When a tool reports that a wafer has passed through a specific chamber, that data is recorded by the ERP in real time along with the chamber ID — this is exactly the wafer-level genealogy fab ERP needs, instead of only tracking aggregated lots like generic manufacturing ERP.
ERP for semiconductor fabs: the "pull" and "push" maintenance-scheduling mechanism against WIP load
Academic research on preventive-maintenance scheduling optimization in semiconductor manufacturing (Yao, Fernández-Gaucherand, Fu, Marcus — Institute for Systems Research, University of Maryland and University of Cincinnati, funded by the Semiconductor Research Corporation and Intel) explicitly models the interaction between PM schedules and WIP (Work-In-Progress): PM should be avoided where possible during periods when a significant amount of work is expected to arrive soon, and a scheduled PM should be "pulled" forward or "pushed" past a certain window when that situation arises.
The "push/defer" rule under high WIP load
When a station is facing a large queue of wafers waiting to be processed and the production target is at risk, the system needs to check a constraint: if this tool is pulled from production for maintenance right now, can the remaining tools at the station still meet the target output? If the answer is no, that tool's PM schedule should be pushed back — even if it's nearly due — to prioritize clearing the WIP first. This rule matters especially for tools assigned to high-priority lots (hot lots).
The "pull/bring forward" rule when a flow gap opens up
Conversely, when the fab is running at high capacity and a tool's wear rate accelerates faster than expected, or when a natural gap opens up in the WIP flow, the system should exploit it to bring PM forward from its original static schedule — inserting downtime at the moment least disruptive to overall flow, instead of letting it land during a high-load period.
Why this is an ERP problem, not a pure CMMS problem
The CMMS knows which tool is nearing its maintenance threshold based on physical wear counters (run hours, wafers processed). But only the ERP knows the full picture of production targets, priority lots, and expected station-level workload over the coming days. The "push" or "pull" decision on a PM schedule can only be made correctly when these two data streams — physical wear from the CMMS and workload from the ERP — are cross-checked together in real time, not handled separately by two disconnected systems.
Illustrative scenario: avoiding a broken production plan from a rigid PM schedule
This is an illustrative scenario for a common type of problem in the industry, not a specific case from any named fab: an etch station runs three tools in parallel, one of which is nearing preventive maintenance under a fixed static schedule. At exactly that moment, the station is handling a surge of high-priority lots, and pulling one of the three tools for maintenance would leave the other two unable to meet the week's production target. If the system only follows the static PM schedule without checking the WIP-load constraint, the production plan breaks. Once the ERP integrates this constraint check before confirming the PM schedule, the decision to push the PM back a few days (within safe wear-limit bounds) is made automatically, and the maintenance is rescheduled to a lower-WIP window.
Reference table: GEM300 component — role — ERP data needed — system link
| Component | Role | ERP data needed | System link |
|---|---|---|---|
| SEMI E87 (Carrier Management) | Automatic FOUP carrier authentication | Carrier status, position within the OHT system | Real-time sync with the tool |
| SEMI E90 (Substrate Tracking) | Tracks each wafer's position/history | Wafer genealogy per processing chamber | Event logged the moment a wafer passes a chamber |
| SEMI E94 (Control Job Management) | Converts production plan into automated commands | Production plan, lot priority | Commands sent down to station controller/host |
| PM-WIP scheduling (pull/push) | Coordinates maintenance schedule against workload | Wear counters (from CMMS) + projected WIP load (from ERP) | Real-time cross-check between CMMS and ERP |
Conclusion
"A good fab ERP isn't the one that follows the maintenance schedule most punctually — it's the one that knows when punctuality would break the production plan, and when to exploit a gap to run maintenance early instead."
Three things worth doing this week if you are evaluating or running an ERP at a semiconductor fab:
- Check whether the system integrates GEM300 (E87/E90/E94) for real-time wafer genealogy, or still manages by aggregated lot like generic manufacturing ERP.
- Confirm whether the PM schedule is checked against WIP-load constraints before confirmation, or just follows a fixed static schedule.
- Review a few recent PM events: has a tool ever been pulled for maintenance right when its station was drowning in hot lots? If so, that's the gap to close between CMMS and ERP.